net/rds/send.c

   1 /*
   2  * Copyright (c) 2006, 2018 Oracle and/or its affiliates. All rights reserved.
   3  *
   4  * This software is available to you under a choice of one of two
   5  * licenses.  You may choose to be licensed under the terms of the GNU
   6  * General Public License (GPL) Version 2, available from the file
   7  * COPYING in the main directory of this source tree, or the
   8  * OpenIB.org BSD license below:
   9  *
  10  *     Redistribution and use in source and binary forms, with or
  11  *     without modification, are permitted provided that the following
  12  *     conditions are met:
  13  *
  14  *      - Redistributions of source code must retain the above
  15  *        copyright notice, this list of conditions and the following
  16  *        disclaimer.
  17  *
  18  *      - Redistributions in binary form must reproduce the above
  19  *        copyright notice, this list of conditions and the following
  20  *        disclaimer in the documentation and/or other materials
  21  *        provided with the distribution.
  22  *
  23  * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
  24  * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
  25  * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
  26  * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
  27  * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
  28  * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
  29  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  30  * SOFTWARE.
  31  *
  32  */
  33 #include <linux/kernel.h>
  34 #include <linux/moduleparam.h>
  35 #include <linux/gfp.h>
  36 #include <net/sock.h>
  37 #include <linux/in.h>
  38 #include <linux/list.h>
  39 #include <linux/ratelimit.h>
  40 #include <linux/export.h>
  41 #include <linux/sizes.h>
  42
  43 #include "rds.h"
  44
  45 /* When transmitting messages in rds_send_xmit, we need to emerge from
  46  * time to time and briefly release the CPU. Otherwise the softlock watchdog
  47  * will kick our shin.
  48  * Also, it seems fairer to not let one busy connection stall all the
  49  * others.
  50  *
  51  * send_batch_count is the number of times we'll loop in send_xmit. Setting
  52  * it to 0 will restore the old behavior (where we looped until we had
  53  * drained the queue).
  54  */
  55 static int send_batch_count = SZ_1K;
  56 module_param(send_batch_count, int, 0444);
  57 MODULE_PARM_DESC(send_batch_count, " batch factor when working the send queue");
  58
  59 static void rds_send_remove_from_sock(struct list_head *messages, int status);
  60
  61 /*
  62  * Reset the send state.  Callers must ensure that this doesn't race with
  63  * rds_send_xmit().
  64  */
  65 void rds_send_path_reset(struct rds_conn_path *cp)
  66 {
  67         struct rds_message *rm, *tmp;
  68         unsigned long flags;
  69
  70         if (cp->cp_xmit_rm) {
  71                 rm = cp->cp_xmit_rm;
  72                 cp->cp_xmit_rm = NULL;
  73                 /* Tell the user the RDMA op is no longer mapped by the
  74                  * transport. This isn't entirely true (it's flushed out
  75                  * independently) but as the connection is down, there's
  76                  * no ongoing RDMA to/from that memory */
  77                 rds_message_unmapped(rm);
  78                 rds_message_put(rm);
  79         }
  80
  81         cp->cp_xmit_sg = 0;
  82         cp->cp_xmit_hdr_off = 0;
  83         cp->cp_xmit_data_off = 0;
  84         cp->cp_xmit_atomic_sent = 0;
  85         cp->cp_xmit_rdma_sent = 0;
  86         cp->cp_xmit_data_sent = 0;
  87
  88         cp->cp_conn->c_map_queued = 0;
  89
  90         cp->cp_unacked_packets = rds_sysctl_max_unacked_packets;
  91         cp->cp_unacked_bytes = rds_sysctl_max_unacked_bytes;
  92
  93         /* Mark messages as retransmissions, and move them to the send q */
  94         spin_lock_irqsave(&cp->cp_lock, flags);
  95         list_for_each_entry_safe(rm, tmp, &cp->cp_retrans, m_conn_item) {
  96                 set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags);
  97                 set_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags);
  98         }
  99         list_splice_init(&cp->cp_retrans, &cp->cp_send_queue);
 100         spin_unlock_irqrestore(&cp->cp_lock, flags);
 101 }
 102 EXPORT_SYMBOL_GPL(rds_send_path_reset);
 103
 104 static int acquire_in_xmit(struct rds_conn_path *cp)
 105 {
 106         return test_and_set_bit(RDS_IN_XMIT, &cp->cp_flags) == 0;
 107 }
 108
 109 static void release_in_xmit(struct rds_conn_path *cp)
 110 {
 111         clear_bit(RDS_IN_XMIT, &cp->cp_flags);
 112         smp_mb__after_atomic();
 113         /*
 114          * We don't use wait_on_bit()/wake_up_bit() because our waking is in a
 115          * hot path and finding waiters is very rare.  We don't want to walk
 116          * the system-wide hashed waitqueue buckets in the fast path only to
 117          * almost never find waiters.
 118          */
 119         if (waitqueue_active(&cp->cp_waitq))
 120                 wake_up_all(&cp->cp_waitq);
 121 }
 122
 123 /*
 124  * We're making the conscious trade-off here to only send one message
 125  * down the connection at a time.
 126  *   Pro:
 127  *      - tx queueing is a simple fifo list
 128  *      - reassembly is optional and easily done by transports per conn
 129  *      - no per flow rx lookup at all, straight to the socket
 130  *      - less per-frag memory and wire overhead
 131  *   Con:
 132  *      - queued acks can be delayed behind large messages
 133  *   Depends:
 134  *      - small message latency is higher behind queued large messages
 135  *      - large message latency isn't starved by intervening small sends
 136  */
 137 int rds_send_xmit(struct rds_conn_path *cp)
 138 {
 139         struct rds_connection *conn = cp->cp_conn;
 140         struct rds_message *rm;
 141         unsigned long flags;
 142         unsigned int tmp;
 143         struct scatterlist *sg;
 144         int ret = 0;
 145         LIST_HEAD(to_be_dropped);
 146         int batch_count;
 147         unsigned long send_gen = 0;
 148
 149 restart:
 150         batch_count = 0;
 151
 152         /*
 153          * sendmsg calls here after having queued its message on the send
 154          * queue.  We only have one task feeding the connection at a time.  If
 155          * another thread is already feeding the queue then we back off.  This
 156          * avoids blocking the caller and trading per-connection data between
 157          * caches per message.
 158          */
 159         if (!acquire_in_xmit(cp)) {
 160                 rds_stats_inc(s_send_lock_contention);
 161                 ret = -ENOMEM;
 162                 goto out;
 163         }
 164
 165         if (rds_destroy_pending(cp->cp_conn)) {
 166                 release_in_xmit(cp);
 167                 ret = -ENETUNREACH; /* dont requeue send work */
 168                 goto out;
 169         }
 170
 171         /*
 172          * we record the send generation after doing the xmit acquire.
 173          * if someone else manages to jump in and do some work, we'll use
 174          * this to avoid a goto restart farther down.
 175          *
 176          * The acquire_in_xmit() check above ensures that only one
 177          * caller can increment c_send_gen at any time.
 178          */
 179         send_gen = READ_ONCE(cp->cp_send_gen) + 1;
 180         WRITE_ONCE(cp->cp_send_gen, send_gen);
 181
 182         /*
 183          * rds_conn_shutdown() sets the conn state and then tests RDS_IN_XMIT,
 184          * we do the opposite to avoid races.
 185          */
 186         if (!rds_conn_path_up(cp)) {
 187                 release_in_xmit(cp);
 188                 ret = 0;
 189                 goto out;
 190         }
 191
 192         if (conn->c_trans->xmit_path_prepare)
 193                 conn->c_trans->xmit_path_prepare(cp);
 194
 195         /*
 196          * spin trying to push headers and data down the connection until
 197          * the connection doesn't make forward progress.
 198          */
 199         while (1) {
 200
 201                 rm = cp->cp_xmit_rm;
 202
 203                 /*
 204                  * If between sending messages, we can send a pending congestion
 205                  * map update.
 206                  */
 207                 if (!rm && test_and_clear_bit(0, &conn->c_map_queued)) {
 208                         rm = rds_cong_update_alloc(conn);
 209                         if (IS_ERR(rm)) {
 210                                 ret = PTR_ERR(rm);
 211                                 break;
 212                         }
 213                         rm->data.op_active = 1;
 214                         rm->m_inc.i_conn_path = cp;
 215                         rm->m_inc.i_conn = cp->cp_conn;
 216
 217                         cp->cp_xmit_rm = rm;
 218                 }
 219
 220                 /*
 221                  * If not already working on one, grab the next message.
 222                  *
 223                  * cp_xmit_rm holds a ref while we're sending this message down
 224                  * the connction.  We can use this ref while holding the
 225                  * send_sem.. rds_send_reset() is serialized with it.
 226                  */
 227                 if (!rm) {
 228                         unsigned int len;
 229
 230                         batch_count++;
 231
 232                         /* we want to process as big a batch as we can, but
 233                          * we also want to avoid softlockups.  If we've been
 234                          * through a lot of messages, lets back off and see
 235                          * if anyone else jumps in
 236                          */
 237                         if (batch_count >= send_batch_count)
 238                                 goto over_batch;
 239
 240                         spin_lock_irqsave(&cp->cp_lock, flags);
 241
 242                         if (!list_empty(&cp->cp_send_queue)) {
 243                                 rm = list_entry(cp->cp_send_queue.next,
 244                                                 struct rds_message,
 245                                                 m_conn_item);
 246                                 rds_message_addref(rm);
 247
 248                                 /*
 249                                  * Move the message from the send queue to the retransmit
 250                                  * list right away.
 251                                  */
 252                                 list_move_tail(&rm->m_conn_item,
 253                                                &cp->cp_retrans);
 254                         }
 255
 256                         spin_unlock_irqrestore(&cp->cp_lock, flags);
 257
 258                         if (!rm)
 259                                 break;
 260
 261                         /* Unfortunately, the way Infiniband deals with
 262                          * RDMA to a bad MR key is by moving the entire
 263                          * queue pair to error state. We cold possibly
 264                          * recover from that, but right now we drop the
 265                          * connection.
 266                          * Therefore, we never retransmit messages with RDMA ops.
 267                          */
 268                         if (test_bit(RDS_MSG_FLUSH, &rm->m_flags) ||
 269                             (rm->rdma.op_active &&
 270                             test_bit(RDS_MSG_RETRANSMITTED, &rm->m_flags))) {
 271                                 spin_lock_irqsave(&cp->cp_lock, flags);
 272                                 if (test_and_clear_bit(RDS_MSG_ON_CONN, &rm->m_flags))
 273                                         list_move(&rm->m_conn_item, &to_be_dropped);
 274                                 spin_unlock_irqrestore(&cp->cp_lock, flags);
 275                                 continue;
 276                         }
 277
 278                         /* Require an ACK every once in a while */
 279                         len = ntohl(rm->m_inc.i_hdr.h_len);
 280                         if (cp->cp_unacked_packets == 0 ||
 281                             cp->cp_unacked_bytes < len) {
 282                                 set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags);
 283
 284                                 cp->cp_unacked_packets =
 285                                         rds_sysctl_max_unacked_packets;
 286                                 cp->cp_unacked_bytes =
 287                                         rds_sysctl_max_unacked_bytes;
 288                                 rds_stats_inc(s_send_ack_required);
 289                         } else {
 290                                 cp->cp_unacked_bytes -= len;
 291                                 cp->cp_unacked_packets--;
 292                         }
 293
 294                         cp->cp_xmit_rm = rm;
 295                 }
 296
 297                 /* The transport either sends the whole rdma or none of it */
 298                 if (rm->rdma.op_active && !cp->cp_xmit_rdma_sent) {
 299                         rm->m_final_op = &rm->rdma;
 300                         /* The transport owns the mapped memory for now.
 301                          * You can't unmap it while it's on the send queue
 302                          */
 303                         set_bit(RDS_MSG_MAPPED, &rm->m_flags);
 304                         ret = conn->c_trans->xmit_rdma(conn, &rm->rdma);
 305                         if (ret) {
 306                                 clear_bit(RDS_MSG_MAPPED, &rm->m_flags);
 307                                 wake_up_interruptible(&rm->m_flush_wait);
 308                                 break;
 309                         }
 310                         cp->cp_xmit_rdma_sent = 1;
 311
 312                 }
 313
 314                 if (rm->atomic.op_active && !cp->cp_xmit_atomic_sent) {
 315                         rm->m_final_op = &rm->atomic;
 316                         /* The transport owns the mapped memory for now.
 317                          * You can't unmap it while it's on the send queue
 318                          */
 319                         set_bit(RDS_MSG_MAPPED, &rm->m_flags);
 320                         ret = conn->c_trans->xmit_atomic(conn, &rm->atomic);
 321                         if (ret) {
 322                                 clear_bit(RDS_MSG_MAPPED, &rm->m_flags);
 323                                 wake_up_interruptible(&rm->m_flush_wait);
 324                                 break;
 325                         }
 326                         cp->cp_xmit_atomic_sent = 1;
 327
 328                 }
 329
 330                 /*
 331                  * A number of cases require an RDS header to be sent
 332                  * even if there is no data.
 333                  * We permit 0-byte sends; rds-ping depends on this.
 334                  * However, if there are exclusively attached silent ops,
 335                  * we skip the hdr/data send, to enable silent operation.
 336                  */
 337                 if (rm->data.op_nents == 0) {
 338                         int ops_present;
 339                         int all_ops_are_silent = 1;
 340
 341                         ops_present = (rm->atomic.op_active || rm->rdma.op_active);
 342                         if (rm->atomic.op_active && !rm->atomic.op_silent)
 343                                 all_ops_are_silent = 0;
 344                         if (rm->rdma.op_active && !rm->rdma.op_silent)
 345                                 all_ops_are_silent = 0;
 346
 347                         if (ops_present && all_ops_are_silent
 348                             && !rm->m_rdma_cookie)
 349                                 rm->data.op_active = 0;
 350                 }
 351
 352                 if (rm->data.op_active && !cp->cp_xmit_data_sent) {
 353                         rm->m_final_op = &rm->data;
 354
 355                         ret = conn->c_trans->xmit(conn, rm,
 356                                                   cp->cp_xmit_hdr_off,
 357                                                   cp->cp_xmit_sg,
 358                                                   cp->cp_xmit_data_off);
 359                         if (ret <= 0)
 360                                 break;
 361
 362                         if (cp->cp_xmit_hdr_off < sizeof(struct rds_header)) {
 363                                 tmp = min_t(int, ret,
 364                                             sizeof(struct rds_header) -
 365                                             cp->cp_xmit_hdr_off);
 366                                 cp->cp_xmit_hdr_off += tmp;
 367                                 ret -= tmp;
 368                         }
 369
 370                         sg = &rm->data.op_sg[cp->cp_xmit_sg];
 371                         while (ret) {
 372                                 tmp = min_t(int, ret, sg->length -
 373                                                       cp->cp_xmit_data_off);
 374                                 cp->cp_xmit_data_off += tmp;
 375                                 ret -= tmp;
 376                                 if (cp->cp_xmit_data_off == sg->length) {
 377                                         cp->cp_xmit_data_off = 0;
 378                                         sg++;
 379                                         cp->cp_xmit_sg++;
 380                                         BUG_ON(ret != 0 && cp->cp_xmit_sg ==
 381                                                rm->data.op_nents);
 382                                 }
 383                         }
 384
 385                         if (cp->cp_xmit_hdr_off == sizeof(struct rds_header) &&
 386                             (cp->cp_xmit_sg == rm->data.op_nents))
 387                                 cp->cp_xmit_data_sent = 1;
 388                 }
 389
 390                 /*
 391                  * A rm will only take multiple times through this loop
 392                  * if there is a data op. Thus, if the data is sent (or there was
 393                  * none), then we're done with the rm.
 394                  */
 395                 if (!rm->data.op_active || cp->cp_xmit_data_sent) {
 396                         cp->cp_xmit_rm = NULL;
 397                         cp->cp_xmit_sg = 0;
 398                         cp->cp_xmit_hdr_off = 0;
 399                         cp->cp_xmit_data_off = 0;
 400                         cp->cp_xmit_rdma_sent = 0;
 401                         cp->cp_xmit_atomic_sent = 0;
 402                         cp->cp_xmit_data_sent = 0;
 403
 404                         rds_message_put(rm);
 405                 }
 406         }
 407
 408 over_batch:
 409         if (conn->c_trans->xmit_path_complete)
 410                 conn->c_trans->xmit_path_complete(cp);
 411         release_in_xmit(cp);
 412
 413         /* Nuke any messages we decided not to retransmit. */
 414         if (!list_empty(&to_be_dropped)) {
 415                 /* irqs on here, so we can put(), unlike above */
 416                 list_for_each_entry(rm, &to_be_dropped, m_conn_item)
 417                         rds_message_put(rm);
 418                 rds_send_remove_from_sock(&to_be_dropped, RDS_RDMA_DROPPED);
 419         }
 420
 421         /*
 422          * Other senders can queue a message after we last test the send queue
 423          * but before we clear RDS_IN_XMIT.  In that case they'd back off and
 424          * not try and send their newly queued message.  We need to check the
 425          * send queue after having cleared RDS_IN_XMIT so that their message
 426          * doesn't get stuck on the send queue.
 427          *
 428          * If the transport cannot continue (i.e ret != 0), then it must
 429          * call us when more room is available, such as from the tx
 430          * completion handler.
 431          *
 432          * We have an extra generation check here so that if someone manages
 433          * to jump in after our release_in_xmit, we'll see that they have done
 434          * some work and we will skip our goto
 435          */
 436         if (ret == 0) {
 437                 bool raced;
 438
 439                 smp_mb();
 440                 raced = send_gen != READ_ONCE(cp->cp_send_gen);
 441
 442                 if ((test_bit(0, &conn->c_map_queued) ||
 443                     !list_empty(&cp->cp_send_queue)) && !raced) {
 444                         if (batch_count < send_batch_count)
 445                                 goto restart;
 446                         rcu_read_lock();
 447                         if (rds_destroy_pending(cp->cp_conn))
 448                                 ret = -ENETUNREACH;
 449                         else
 450                                 queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
 451                         rcu_read_unlock();
 452                 } else if (raced) {
 453                         rds_stats_inc(s_send_lock_queue_raced);
 454                 }
 455         }
 456 out:
 457         return ret;
 458 }
 459 EXPORT_SYMBOL_GPL(rds_send_xmit);
 460
 461 static void rds_send_sndbuf_remove(struct rds_sock *rs, struct rds_message *rm)
 462 {
 463         u32 len = be32_to_cpu(rm->m_inc.i_hdr.h_len);
 464
 465         assert_spin_locked(&rs->rs_lock);
 466
 467         BUG_ON(rs->rs_snd_bytes < len);
 468         rs->rs_snd_bytes -= len;
 469
 470         if (rs->rs_snd_bytes == 0)
 471                 rds_stats_inc(s_send_queue_empty);
 472 }
 473
 474 static inline int rds_send_is_acked(struct rds_message *rm, u64 ack,
 475                                     is_acked_func is_acked)
 476 {
 477         if (is_acked)
 478                 return is_acked(rm, ack);
 479         return be64_to_cpu(rm->m_inc.i_hdr.h_sequence) <= ack;
 480 }
 481
 482 /*
 483  * This is pretty similar to what happens below in the ACK
 484  * handling code - except that we call here as soon as we get
 485  * the IB send completion on the RDMA op and the accompanying
 486  * message.
 487  */
 488 void rds_rdma_send_complete(struct rds_message *rm, int status)
 489 {
 490         struct rds_sock *rs = NULL;
 491         struct rm_rdma_op *ro;
 492         struct rds_notifier *notifier;
 493         unsigned long flags;
 494
 495         spin_lock_irqsave(&rm->m_rs_lock, flags);
 496
 497         ro = &rm->rdma;
 498         if (test_bit(RDS_MSG_ON_SOCK, &rm->m_flags) &&
 499             ro->op_active && ro->op_notify && ro->op_notifier) {
 500                 notifier = ro->op_notifier;
 501                 rs = rm->m_rs;
 502                 sock_hold(rds_rs_to_sk(rs));
 503
 504                 notifier->n_status = status;
 505                 spin_lock(&rs->rs_lock);
 506                 list_add_tail(&notifier->n_list, &rs->rs_notify_queue);
 507                 spin_unlock(&rs->rs_lock);
 508
 509                 ro->op_notifier = NULL;
 510         }
 511
 512         spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 513
 514         if (rs) {
 515                 rds_wake_sk_sleep(rs);
 516                 sock_put(rds_rs_to_sk(rs));
 517         }
 518 }
 519 EXPORT_SYMBOL_GPL(rds_rdma_send_complete);
 520
 521 /*
 522  * Just like above, except looks at atomic op
 523  */
 524 void rds_atomic_send_complete(struct rds_message *rm, int status)
 525 {
 526         struct rds_sock *rs = NULL;
 527         struct rm_atomic_op *ao;
 528         struct rds_notifier *notifier;
 529         unsigned long flags;
 530
 531         spin_lock_irqsave(&rm->m_rs_lock, flags);
 532
 533         ao = &rm->atomic;
 534         if (test_bit(RDS_MSG_ON_SOCK, &rm->m_flags)
 535             && ao->op_active && ao->op_notify && ao->op_notifier) {
 536                 notifier = ao->op_notifier;
 537                 rs = rm->m_rs;
 538                 sock_hold(rds_rs_to_sk(rs));
 539
 540                 notifier->n_status = status;
 541                 spin_lock(&rs->rs_lock);
 542                 list_add_tail(&notifier->n_list, &rs->rs_notify_queue);
 543                 spin_unlock(&rs->rs_lock);
 544
 545                 ao->op_notifier = NULL;
 546         }
 547
 548         spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 549
 550         if (rs) {
 551                 rds_wake_sk_sleep(rs);
 552                 sock_put(rds_rs_to_sk(rs));
 553         }
 554 }
 555 EXPORT_SYMBOL_GPL(rds_atomic_send_complete);
 556
 557 /*
 558  * This is the same as rds_rdma_send_complete except we
 559  * don't do any locking - we have all the ingredients (message,
 560  * socket, socket lock) and can just move the notifier.
 561  */
 562 static inline void
 563 __rds_send_complete(struct rds_sock *rs, struct rds_message *rm, int status)
 564 {
 565         struct rm_rdma_op *ro;
 566         struct rm_atomic_op *ao;
 567
 568         ro = &rm->rdma;
 569         if (ro->op_active && ro->op_notify && ro->op_notifier) {
 570                 ro->op_notifier->n_status = status;
 571                 list_add_tail(&ro->op_notifier->n_list, &rs->rs_notify_queue);
 572                 ro->op_notifier = NULL;
 573         }
 574
 575         ao = &rm->atomic;
 576         if (ao->op_active && ao->op_notify && ao->op_notifier) {
 577                 ao->op_notifier->n_status = status;
 578                 list_add_tail(&ao->op_notifier->n_list, &rs->rs_notify_queue);
 579                 ao->op_notifier = NULL;
 580         }
 581
 582         /* No need to wake the app - caller does this */
 583 }
 584
 585 /*
 586  * This removes messages from the socket's list if they're on it.  The list
 587  * argument must be private to the caller, we must be able to modify it
 588  * without locks.  The messages must have a reference held for their
 589  * position on the list.  This function will drop that reference after
 590  * removing the messages from the 'messages' list regardless of if it found
 591  * the messages on the socket list or not.
 592  */
 593 static void rds_send_remove_from_sock(struct list_head *messages, int status)
 594 {
 595         unsigned long flags;
 596         struct rds_sock *rs = NULL;
 597         struct rds_message *rm;
 598
 599         while (!list_empty(messages)) {
 600                 int was_on_sock = 0;
 601
 602                 rm = list_entry(messages->next, struct rds_message,
 603                                 m_conn_item);
 604                 list_del_init(&rm->m_conn_item);
 605
 606                 /*
 607                  * If we see this flag cleared then we're *sure* that someone
 608                  * else beat us to removing it from the sock.  If we race
 609                  * with their flag update we'll get the lock and then really
 610                  * see that the flag has been cleared.
 611                  *
 612                  * The message spinlock makes sure nobody clears rm->m_rs
 613                  * while we're messing with it. It does not prevent the
 614                  * message from being removed from the socket, though.
 615                  */
 616                 spin_lock_irqsave(&rm->m_rs_lock, flags);
 617                 if (!test_bit(RDS_MSG_ON_SOCK, &rm->m_flags))
 618                         goto unlock_and_drop;
 619
 620                 if (rs != rm->m_rs) {
 621                         if (rs) {
 622                                 rds_wake_sk_sleep(rs);
 623                                 sock_put(rds_rs_to_sk(rs));
 624                         }
 625                         rs = rm->m_rs;
 626                         if (rs)
 627                                 sock_hold(rds_rs_to_sk(rs));
 628                 }
 629                 if (!rs)
 630                         goto unlock_and_drop;
 631                 spin_lock(&rs->rs_lock);
 632
 633                 if (test_and_clear_bit(RDS_MSG_ON_SOCK, &rm->m_flags)) {
 634                         struct rm_rdma_op *ro = &rm->rdma;
 635                         struct rds_notifier *notifier;
 636
 637                         list_del_init(&rm->m_sock_item);
 638                         rds_send_sndbuf_remove(rs, rm);
 639
 640                         if (ro->op_active && ro->op_notifier &&
 641                                (ro->op_notify || (ro->op_recverr && status))) {
 642                                 notifier = ro->op_notifier;
 643                                 list_add_tail(&notifier->n_list,
 644                                                 &rs->rs_notify_queue);
 645                                 if (!notifier->n_status)
 646                                         notifier->n_status = status;
 647                                 rm->rdma.op_notifier = NULL;
 648                         }
 649                         was_on_sock = 1;
 650                 }
 651                 spin_unlock(&rs->rs_lock);
 652
 653 unlock_and_drop:
 654                 spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 655                 rds_message_put(rm);
 656                 if (was_on_sock)
 657                         rds_message_put(rm);
 658         }
 659
 660         if (rs) {
 661                 rds_wake_sk_sleep(rs);
 662                 sock_put(rds_rs_to_sk(rs));
 663         }
 664 }
 665
 666 /*
 667  * Transports call here when they've determined that the receiver queued
 668  * messages up to, and including, the given sequence number.  Messages are
 669  * moved to the retrans queue when rds_send_xmit picks them off the send
 670  * queue. This means that in the TCP case, the message may not have been
 671  * assigned the m_ack_seq yet - but that's fine as long as tcp_is_acked
 672  * checks the RDS_MSG_HAS_ACK_SEQ bit.
 673  */
 674 void rds_send_path_drop_acked(struct rds_conn_path *cp, u64 ack,
 675                               is_acked_func is_acked)
 676 {
 677         struct rds_message *rm, *tmp;
 678         unsigned long flags;
 679         LIST_HEAD(list);
 680
 681         spin_lock_irqsave(&cp->cp_lock, flags);
 682
 683         list_for_each_entry_safe(rm, tmp, &cp->cp_retrans, m_conn_item) {
 684                 if (!rds_send_is_acked(rm, ack, is_acked))
 685                         break;
 686
 687                 list_move(&rm->m_conn_item, &list);
 688                 clear_bit(RDS_MSG_ON_CONN, &rm->m_flags);
 689         }
 690
 691         /* order flag updates with spin locks */
 692         if (!list_empty(&list))
 693                 smp_mb__after_atomic();
 694
 695         spin_unlock_irqrestore(&cp->cp_lock, flags);
 696
 697         /* now remove the messages from the sock list as needed */
 698         rds_send_remove_from_sock(&list, RDS_RDMA_SUCCESS);
 699 }
 700 EXPORT_SYMBOL_GPL(rds_send_path_drop_acked);
 701
 702 void rds_send_drop_acked(struct rds_connection *conn, u64 ack,
 703                          is_acked_func is_acked)
 704 {
 705         WARN_ON(conn->c_trans->t_mp_capable);
 706         rds_send_path_drop_acked(&conn->c_path[0], ack, is_acked);
 707 }
 708 EXPORT_SYMBOL_GPL(rds_send_drop_acked);
 709
 710 void rds_send_drop_to(struct rds_sock *rs, struct sockaddr_in6 *dest)
 711 {
 712         struct rds_message *rm, *tmp;
 713         struct rds_connection *conn;
 714         struct rds_conn_path *cp;
 715         unsigned long flags;
 716         LIST_HEAD(list);
 717
 718         /* get all the messages we're dropping under the rs lock */
 719         spin_lock_irqsave(&rs->rs_lock, flags);
 720
 721         list_for_each_entry_safe(rm, tmp, &rs->rs_send_queue, m_sock_item) {
 722                 if (dest &&
 723                     (!ipv6_addr_equal(&dest->sin6_addr, &rm->m_daddr) ||
 724                      dest->sin6_port != rm->m_inc.i_hdr.h_dport))
 725                         continue;
 726
 727                 list_move(&rm->m_sock_item, &list);
 728                 rds_send_sndbuf_remove(rs, rm);
 729                 clear_bit(RDS_MSG_ON_SOCK, &rm->m_flags);
 730         }
 731
 732         /* order flag updates with the rs lock */
 733         smp_mb__after_atomic();
 734
 735         spin_unlock_irqrestore(&rs->rs_lock, flags);
 736
 737         if (list_empty(&list))
 738                 return;
 739
 740         /* Remove the messages from the conn */
 741         list_for_each_entry(rm, &list, m_sock_item) {
 742
 743                 conn = rm->m_inc.i_conn;
 744                 if (conn->c_trans->t_mp_capable)
 745                         cp = rm->m_inc.i_conn_path;
 746                 else
 747                         cp = &conn->c_path[0];
 748
 749                 spin_lock_irqsave(&cp->cp_lock, flags);
 750                 /*
 751                  * Maybe someone else beat us to removing rm from the conn.
 752                  * If we race with their flag update we'll get the lock and
 753                  * then really see that the flag has been cleared.
 754                  */
 755                 if (!test_and_clear_bit(RDS_MSG_ON_CONN, &rm->m_flags)) {
 756                         spin_unlock_irqrestore(&cp->cp_lock, flags);
 757                         continue;
 758                 }
 759                 list_del_init(&rm->m_conn_item);
 760                 spin_unlock_irqrestore(&cp->cp_lock, flags);
 761
 762                 /*
 763                  * Couldn't grab m_rs_lock in top loop (lock ordering),
 764                  * but we can now.
 765                  */
 766                 spin_lock_irqsave(&rm->m_rs_lock, flags);
 767
 768                 spin_lock(&rs->rs_lock);
 769                 __rds_send_complete(rs, rm, RDS_RDMA_CANCELED);
 770                 spin_unlock(&rs->rs_lock);
 771
 772                 spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 773
 774                 rds_message_put(rm);
 775         }
 776
 777         rds_wake_sk_sleep(rs);
 778
 779         while (!list_empty(&list)) {
 780                 rm = list_entry(list.next, struct rds_message, m_sock_item);
 781                 list_del_init(&rm->m_sock_item);
 782                 rds_message_wait(rm);
 783
 784                 /* just in case the code above skipped this message
 785                  * because RDS_MSG_ON_CONN wasn't set, run it again here
 786                  * taking m_rs_lock is the only thing that keeps us
 787                  * from racing with ack processing.
 788                  */
 789                 spin_lock_irqsave(&rm->m_rs_lock, flags);
 790
 791                 spin_lock(&rs->rs_lock);
 792                 __rds_send_complete(rs, rm, RDS_RDMA_CANCELED);
 793                 spin_unlock(&rs->rs_lock);
 794
 795                 spin_unlock_irqrestore(&rm->m_rs_lock, flags);
 796
 797                 rds_message_put(rm);
 798         }
 799 }
 800
 801 /*
 802  * we only want this to fire once so we use the callers 'queued'.  It's
 803  * possible that another thread can race with us and remove the
 804  * message from the flow with RDS_CANCEL_SENT_TO.
 805  */
 806 static int rds_send_queue_rm(struct rds_sock *rs, struct rds_connection *conn,
 807                              struct rds_conn_path *cp,
 808                              struct rds_message *rm, __be16 sport,
 809                              __be16 dport, int *queued)
 810 {
 811         unsigned long flags;
 812         u32 len;
 813
 814         if (*queued)
 815                 goto out;
 816
 817         len = be32_to_cpu(rm->m_inc.i_hdr.h_len);
 818
 819         /* this is the only place which holds both the socket's rs_lock
 820          * and the connection's c_lock */
 821         spin_lock_irqsave(&rs->rs_lock, flags);
 822
 823         /*
 824          * If there is a little space in sndbuf, we don't queue anything,
 825          * and userspace gets -EAGAIN. But poll() indicates there's send
 826          * room. This can lead to bad behavior (spinning) if snd_bytes isn't
 827          * freed up by incoming acks. So we check the *old* value of
 828          * rs_snd_bytes here to allow the last msg to exceed the buffer,
 829          * and poll() now knows no more data can be sent.
 830          */
 831         if (rs->rs_snd_bytes < rds_sk_sndbuf(rs)) {
 832                 rs->rs_snd_bytes += len;
 833
 834                 /* let recv side know we are close to send space exhaustion.
 835                  * This is probably not the optimal way to do it, as this
 836                  * means we set the flag on *all* messages as soon as our
 837                  * throughput hits a certain threshold.
 838                  */
 839                 if (rs->rs_snd_bytes >= rds_sk_sndbuf(rs) / 2)
 840                         set_bit(RDS_MSG_ACK_REQUIRED, &rm->m_flags);
 841
 842                 list_add_tail(&rm->m_sock_item, &rs->rs_send_queue);
 843                 set_bit(RDS_MSG_ON_SOCK, &rm->m_flags);
 844                 rds_message_addref(rm);
 845                 sock_hold(rds_rs_to_sk(rs));
 846                 rm->m_rs = rs;
 847
 848                 /* The code ordering is a little weird, but we're
 849                    trying to minimize the time we hold c_lock */
 850                 rds_message_populate_header(&rm->m_inc.i_hdr, sport, dport, 0);
 851                 rm->m_inc.i_conn = conn;
 852                 rm->m_inc.i_conn_path = cp;
 853                 rds_message_addref(rm);
 854
 855                 spin_lock(&cp->cp_lock);
 856                 rm->m_inc.i_hdr.h_sequence = cpu_to_be64(cp->cp_next_tx_seq++);
 857                 list_add_tail(&rm->m_conn_item, &cp->cp_send_queue);
 858                 set_bit(RDS_MSG_ON_CONN, &rm->m_flags);
 859                 spin_unlock(&cp->cp_lock);
 860
 861                 rdsdebug("queued msg %p len %d, rs %p bytes %d seq %llu\n",
 862                          rm, len, rs, rs->rs_snd_bytes,
 863                          (unsigned long long)be64_to_cpu(rm->m_inc.i_hdr.h_sequence));
 864
 865                 *queued = 1;
 866         }
 867
 868         spin_unlock_irqrestore(&rs->rs_lock, flags);
 869 out:
 870         return *queued;
 871 }
 872
 873 /*
 874  * rds_message is getting to be quite complicated, and we'd like to allocate
 875  * it all in one go. This figures out how big it needs to be up front.
 876  */
 877 static int rds_rm_size(struct msghdr *msg, int num_sgs,
 878                        struct rds_iov_vector_arr *vct)
 879 {
 880         struct cmsghdr *cmsg;
 881         int size = 0;
 882         int cmsg_groups = 0;
 883         int retval;
 884         bool zcopy_cookie = false;
 885         struct rds_iov_vector *iov, *tmp_iov;
 886
 887         if (num_sgs < 0)
 888                 return -EINVAL;
 889
 890         for_each_cmsghdr(cmsg, msg) {
 891                 if (!CMSG_OK(msg, cmsg))
 892                         return -EINVAL;
 893
 894                 if (cmsg->cmsg_level != SOL_RDS)
 895                         continue;
 896
 897                 switch (cmsg->cmsg_type) {
 898                 case RDS_CMSG_RDMA_ARGS:
 899                         if (vct->indx >= vct->len) {
 900                                 vct->len += vct->incr;
 901                                 tmp_iov =
 902                                         krealloc(vct->vec,
 903                                                  vct->len *
 904                                                  sizeof(struct rds_iov_vector),
 905                                                  GFP_KERNEL);
 906                                 if (!tmp_iov) {
 907                                         vct->len -= vct->incr;
 908                                         return -ENOMEM;
 909                                 }
 910                                 vct->vec = tmp_iov;
 911                         }
 912                         iov = &vct->vec[vct->indx];
 913                         memset(iov, 0, sizeof(struct rds_iov_vector));
 914                         vct->indx++;
 915                         cmsg_groups |= 1;
 916                         retval = rds_rdma_extra_size(CMSG_DATA(cmsg), iov);
 917                         if (retval < 0)
 918                                 return retval;
 919                         size += retval;
 920
 921                         break;
 922
 923                 case RDS_CMSG_ZCOPY_COOKIE:
 924                         zcopy_cookie = true;
 925                         /* fall through */
 926
 927                 case RDS_CMSG_RDMA_DEST:
 928                 case RDS_CMSG_RDMA_MAP:
 929                         cmsg_groups |= 2;
 930                         /* these are valid but do no add any size */
 931                         break;
 932
 933                 case RDS_CMSG_ATOMIC_CSWP:
 934                 case RDS_CMSG_ATOMIC_FADD:
 935                 case RDS_CMSG_MASKED_ATOMIC_CSWP:
 936                 case RDS_CMSG_MASKED_ATOMIC_FADD:
 937                         cmsg_groups |= 1;
 938                         size += sizeof(struct scatterlist);
 939                         break;
 940
 941                 default:
 942                         return -EINVAL;
 943                 }
 944
 945         }
 946
 947         if ((msg->msg_flags & MSG_ZEROCOPY) && !zcopy_cookie)
 948                 return -EINVAL;
 949
 950         size += num_sgs * sizeof(struct scatterlist);
 951
 952         /* Ensure (DEST, MAP) are never used with (ARGS, ATOMIC) */
 953         if (cmsg_groups == 3)
 954                 return -EINVAL;
 955
 956         return size;
 957 }
 958
 959 static int rds_cmsg_zcopy(struct rds_sock *rs, struct rds_message *rm,
 960                           struct cmsghdr *cmsg)
 961 {
 962         u32 *cookie;
 963
 964         if (cmsg->cmsg_len < CMSG_LEN(sizeof(*cookie)) ||
 965             !rm->data.op_mmp_znotifier)
 966                 return -EINVAL;
 967         cookie = CMSG_DATA(cmsg);
 968         rm->data.op_mmp_znotifier->z_cookie = *cookie;
 969         return 0;
 970 }
 971
 972 static int rds_cmsg_send(struct rds_sock *rs, struct rds_message *rm,
 973                          struct msghdr *msg, int *allocated_mr,
 974                          struct rds_iov_vector_arr *vct)
 975 {
 976         struct cmsghdr *cmsg;
 977         int ret = 0, ind = 0;
 978
 979         for_each_cmsghdr(cmsg, msg) {
 980                 if (!CMSG_OK(msg, cmsg))
 981                         return -EINVAL;
 982
 983                 if (cmsg->cmsg_level != SOL_RDS)
 984                         continue;
 985
 986                 /* As a side effect, RDMA_DEST and RDMA_MAP will set
 987                  * rm->rdma.m_rdma_cookie and rm->rdma.m_rdma_mr.
 988                  */
 989                 switch (cmsg->cmsg_type) {
 990                 case RDS_CMSG_RDMA_ARGS:
 991                         if (ind >= vct->indx)
 992                                 return -ENOMEM;
 993                         ret = rds_cmsg_rdma_args(rs, rm, cmsg, &vct->vec[ind]);
 994                         ind++;
 995                         break;
 996
 997                 case RDS_CMSG_RDMA_DEST:
 998                         ret = rds_cmsg_rdma_dest(rs, rm, cmsg);
 999                         break;
1000
1001                 case RDS_CMSG_RDMA_MAP:
1002                         ret = rds_cmsg_rdma_map(rs, rm, cmsg);
1003                         if (!ret)
1004                                 *allocated_mr = 1;
1005                         else if (ret == -ENODEV)
1006                                 /* Accommodate the get_mr() case which can fail
1007                                  * if connection isn't established yet.
1008                                  */
1009                                 ret = -EAGAIN;
1010                         break;
1011                 case RDS_CMSG_ATOMIC_CSWP:
1012                 case RDS_CMSG_ATOMIC_FADD:
1013                 case RDS_CMSG_MASKED_ATOMIC_CSWP:
1014                 case RDS_CMSG_MASKED_ATOMIC_FADD:
1015                         ret = rds_cmsg_atomic(rs, rm, cmsg);
1016                         break;
1017
1018                 case RDS_CMSG_ZCOPY_COOKIE:
1019                         ret = rds_cmsg_zcopy(rs, rm, cmsg);
1020                         break;
1021
1022                 default:
1023                         return -EINVAL;
1024                 }
1025
1026                 if (ret)
1027                         break;
1028         }
1029
1030         return ret;
1031 }
1032
1033 static int rds_send_mprds_hash(struct rds_sock *rs,
1034                                struct rds_connection *conn, int nonblock)
1035 {
1036         int hash;
1037
1038         if (conn->c_npaths == 0)
1039                 hash = RDS_MPATH_HASH(rs, RDS_MPATH_WORKERS);
1040         else
1041                 hash = RDS_MPATH_HASH(rs, conn->c_npaths);
1042         if (conn->c_npaths == 0 && hash != 0) {
1043                 rds_send_ping(conn, 0);
1044
1045                 /* The underlying connection is not up yet.  Need to wait
1046                  * until it is up to be sure that the non-zero c_path can be
1047                  * used.  But if we are interrupted, we have to use the zero
1048                  * c_path in case the connection ends up being non-MP capable.
1049                  */
1050                 if (conn->c_npaths == 0) {
1051                         /* Cannot wait for the connection be made, so just use
1052                          * the base c_path.
1053                          */
1054                         if (nonblock)
1055                                 return 0;
1056                         if (wait_event_interruptible(conn->c_hs_waitq,
1057                                                      conn->c_npaths != 0))
1058                                 hash = 0;
1059                 }
1060                 if (conn->c_npaths == 1)
1061                         hash = 0;
1062         }
1063         return hash;
1064 }
1065
1066 static int rds_rdma_bytes(struct msghdr *msg, size_t *rdma_bytes)
1067 {
1068         struct rds_rdma_args *args;
1069         struct cmsghdr *cmsg;
1070
1071         for_each_cmsghdr(cmsg, msg) {
1072                 if (!CMSG_OK(msg, cmsg))
1073                         return -EINVAL;
1074
1075                 if (cmsg->cmsg_level != SOL_RDS)
1076                         continue;
1077
1078                 if (cmsg->cmsg_type == RDS_CMSG_RDMA_ARGS) {
1079                         if (cmsg->cmsg_len <
1080                             CMSG_LEN(sizeof(struct rds_rdma_args)))
1081                                 return -EINVAL;
1082                         args = CMSG_DATA(cmsg);
1083                         *rdma_bytes += args->remote_vec.bytes;
1084                 }
1085         }
1086         return 0;
1087 }
1088
1089 int rds_sendmsg(struct socket *sock, struct msghdr *msg, size_t payload_len)
1090 {
1091         struct sock *sk = sock->sk;
1092         struct rds_sock *rs = rds_sk_to_rs(sk);
1093         DECLARE_SOCKADDR(struct sockaddr_in6 *, sin6, msg->msg_name);
1094         DECLARE_SOCKADDR(struct sockaddr_in *, usin, msg->msg_name);
1095         __be16 dport;
1096         struct rds_message *rm = NULL;
1097         struct rds_connection *conn;
1098         int ret = 0;
1099         int queued = 0, allocated_mr = 0;
1100         int nonblock = msg->msg_flags & MSG_DONTWAIT;
1101         long timeo = sock_sndtimeo(sk, nonblock);
1102         struct rds_conn_path *cpath;
1103         struct in6_addr daddr;
1104         __u32 scope_id = 0;
1105         size_t total_payload_len = payload_len, rdma_payload_len = 0;
1106         bool zcopy = ((msg->msg_flags & MSG_ZEROCOPY) &&
1107                       sock_flag(rds_rs_to_sk(rs), SOCK_ZEROCOPY));
1108         int num_sgs = DIV_ROUND_UP(payload_len, PAGE_SIZE);
1109         int namelen;
1110         struct rds_iov_vector_arr vct;
1111         int ind;
1112
1113         memset(&vct, 0, sizeof(vct));
1114
1115         /* expect 1 RDMA CMSG per rds_sendmsg. can still grow if more needed. */
1116         vct.incr = 1;
1117
1118         /* Mirror Linux UDP mirror of BSD error message compatibility */
1119         /* XXX: Perhaps MSG_MORE someday */
1120         if (msg->msg_flags & ~(MSG_DONTWAIT | MSG_CMSG_COMPAT | MSG_ZEROCOPY)) {
1121                 ret = -EOPNOTSUPP;
1122                 goto out;
1123         }
1124
1125         namelen = msg->msg_namelen;
1126         if (namelen != 0) {
1127                 if (namelen < sizeof(*usin)) {
1128                         ret = -EINVAL;
1129                         goto out;
1130                 }
1131                 switch (usin->sin_family) {
1132                 case AF_INET:
1133                         if (usin->sin_addr.s_addr == htonl(INADDR_ANY) ||
1134                             usin->sin_addr.s_addr == htonl(INADDR_BROADCAST) ||
1135                             IN_MULTICAST(ntohl(usin->sin_addr.s_addr))) {
1136                                 ret = -EINVAL;
1137                                 goto out;
1138                         }
1139                         ipv6_addr_set_v4mapped(usin->sin_addr.s_addr, &daddr);
1140                         dport = usin->sin_port;
1141                         break;
1142
1143 #if IS_ENABLED(CONFIG_IPV6)
1144                 case AF_INET6: {
1145                         int addr_type;
1146
1147                         if (namelen < sizeof(*sin6)) {
1148                                 ret = -EINVAL;
1149                                 goto out;
1150                         }
1151                         addr_type = ipv6_addr_type(&sin6->sin6_addr);
1152                         if (!(addr_type & IPV6_ADDR_UNICAST)) {
1153                                 __be32 addr4;
1154
1155                                 if (!(addr_type & IPV6_ADDR_MAPPED)) {
1156                                         ret = -EINVAL;
1157                                         goto out;
1158                                 }
1159
1160                                 /* It is a mapped address.  Need to do some
1161                                  * sanity checks.
1162                                  */
1163                                 addr4 = sin6->sin6_addr.s6_addr32[3];
1164                                 if (addr4 == htonl(INADDR_ANY) ||
1165                                     addr4 == htonl(INADDR_BROADCAST) ||
1166                                     IN_MULTICAST(ntohl(addr4))) {
1167                                         ret = -EINVAL;
1168                                         goto out;
1169                                 }
1170                         }
1171                         if (addr_type & IPV6_ADDR_LINKLOCAL) {
1172                                 if (sin6->sin6_scope_id == 0) {
1173                                         ret = -EINVAL;
1174                                         goto out;
1175                                 }
1176                                 scope_id = sin6->sin6_scope_id;
1177                         }
1178
1179                         daddr = sin6->sin6_addr;
1180                         dport = sin6->sin6_port;
1181                         break;
1182                 }
1183 #endif
1184
1185                 default:
1186                         ret = -EINVAL;
1187                         goto out;
1188                 }
1189         } else {
1190                 /* We only care about consistency with ->connect() */
1191                 lock_sock(sk);
1192                 daddr = rs->rs_conn_addr;
1193                 dport = rs->rs_conn_port;
1194                 scope_id = rs->rs_bound_scope_id;
1195                 release_sock(sk);
1196         }
1197
1198         lock_sock(sk);
1199         if (ipv6_addr_any(&rs->rs_bound_addr) || ipv6_addr_any(&daddr)) {
1200                 release_sock(sk);
1201                 ret = -ENOTCONN;
1202                 goto out;
1203         } else if (namelen != 0) {
1204                 /* Cannot send to an IPv4 address using an IPv6 source
1205                  * address and cannot send to an IPv6 address using an
1206                  * IPv4 source address.
1207                  */
1208                 if (ipv6_addr_v4mapped(&daddr) ^
1209                     ipv6_addr_v4mapped(&rs->rs_bound_addr)) {
1210                         release_sock(sk);
1211                         ret = -EOPNOTSUPP;
1212                         goto out;
1213                 }
1214                 /* If the socket is already bound to a link local address,
1215                  * it can only send to peers on the same link.  But allow
1216                  * communicating beween link local and non-link local address.
1217                  */
1218                 if (scope_id != rs->rs_bound_scope_id) {
1219                         if (!scope_id) {
1220                                 scope_id = rs->rs_bound_scope_id;
1221                         } else if (rs->rs_bound_scope_id) {
1222                                 release_sock(sk);
1223                                 ret = -EINVAL;
1224                                 goto out;
1225                         }
1226                 }
1227         }
1228         release_sock(sk);
1229
1230         ret = rds_rdma_bytes(msg, &rdma_payload_len);
1231         if (ret)
1232                 goto out;
1233
1234         total_payload_len += rdma_payload_len;
1235         if (max_t(size_t, payload_len, rdma_payload_len) > RDS_MAX_MSG_SIZE) {
1236                 ret = -EMSGSIZE;
1237                 goto out;
1238         }
1239
1240         if (payload_len > rds_sk_sndbuf(rs)) {
1241                 ret = -EMSGSIZE;
1242                 goto out;
1243         }
1244
1245         if (zcopy) {
1246                 if (rs->rs_transport->t_type != RDS_TRANS_TCP) {
1247                         ret = -EOPNOTSUPP;
1248                         goto out;
1249                 }
1250                 num_sgs = iov_iter_npages(&msg->msg_iter, INT_MAX);
1251         }
1252         /* size of rm including all sgs */
1253         ret = rds_rm_size(msg, num_sgs, &vct);
1254         if (ret < 0)
1255                 goto out;
1256
1257         rm = rds_message_alloc(ret, GFP_KERNEL);
1258         if (!rm) {
1259                 ret = -ENOMEM;
1260                 goto out;
1261         }
1262
1263         /* Attach data to the rm */
1264         if (payload_len) {
1265                 rm->data.op_sg = rds_message_alloc_sgs(rm, num_sgs, &ret);
1266                 if (!rm->data.op_sg)
1267                         goto out;
1268                 ret = rds_message_copy_from_user(rm, &msg->msg_iter, zcopy);
1269                 if (ret)
1270                         goto out;
1271         }
1272         rm->data.op_active = 1;
1273
1274         rm->m_daddr = daddr;
1275
1276         /* rds_conn_create has a spinlock that runs with IRQ off.
1277          * Caching the conn in the socket helps a lot. */
1278         if (rs->rs_conn && ipv6_addr_equal(&rs->rs_conn->c_faddr, &daddr) &&
1279             rs->rs_tos == rs->rs_conn->c_tos) {
1280                 conn = rs->rs_conn;
1281         } else {
1282                 conn = rds_conn_create_outgoing(sock_net(sock->sk),
1283                                                 &rs->rs_bound_addr, &daddr,
1284                                                 rs->rs_transport, rs->rs_tos,
1285                                                 sock->sk->sk_allocation,
1286                                                 scope_id);
1287                 if (IS_ERR(conn)) {
1288                         ret = PTR_ERR(conn);
1289                         goto out;
1290                 }
1291                 rs->rs_conn = conn;
1292         }
1293
1294         if (conn->c_trans->t_mp_capable)
1295                 cpath = &conn->c_path[rds_send_mprds_hash(rs, conn, nonblock)];
1296         else
1297                 cpath = &conn->c_path[0];
1298
1299         rm->m_conn_path = cpath;
1300
1301         /* Parse any control messages the user may have included. */
1302         ret = rds_cmsg_send(rs, rm, msg, &allocated_mr, &vct);
1303         if (ret) {
1304                 /* Trigger connection so that its ready for the next retry */
1305                 if (ret ==  -EAGAIN)
1306                         rds_conn_connect_if_down(conn);
1307                 goto out;
1308         }
1309
1310         if (rm->rdma.op_active && !conn->c_trans->xmit_rdma) {
1311                 printk_ratelimited(KERN_NOTICE "rdma_op %p conn xmit_rdma %p\n",
1312                                &rm->rdma, conn->c_trans->xmit_rdma);
1313                 ret = -EOPNOTSUPP;
1314                 goto out;
1315         }
1316
1317         if (rm->atomic.op_active && !conn->c_trans->xmit_atomic) {
1318                 printk_ratelimited(KERN_NOTICE "atomic_op %p conn xmit_atomic %p\n",
1319                                &rm->atomic, conn->c_trans->xmit_atomic);
1320                 ret = -EOPNOTSUPP;
1321                 goto out;
1322         }
1323
1324         if (rds_destroy_pending(conn)) {
1325                 ret = -EAGAIN;
1326                 goto out;
1327         }
1328
1329         rds_conn_path_connect_if_down(cpath);
1330
1331         ret = rds_cong_wait(conn->c_fcong, dport, nonblock, rs);
1332         if (ret) {
1333                 rs->rs_seen_congestion = 1;
1334                 goto out;
1335         }
1336         while (!rds_send_queue_rm(rs, conn, cpath, rm, rs->rs_bound_port,
1337                                   dport, &queued)) {
1338                 rds_stats_inc(s_send_queue_full);
1339
1340                 if (nonblock) {
1341                         ret = -EAGAIN;
1342                         goto out;
1343                 }
1344
1345                 timeo = wait_event_interruptible_timeout(*sk_sleep(sk),
1346                                         rds_send_queue_rm(rs, conn, cpath, rm,
1347                                                           rs->rs_bound_port,
1348                                                           dport,
1349                                                           &queued),
1350                                         timeo);
1351                 rdsdebug("sendmsg woke queued %d timeo %ld\n", queued, timeo);
1352                 if (timeo > 0 || timeo == MAX_SCHEDULE_TIMEOUT)
1353                         continue;
1354
1355                 ret = timeo;
1356                 if (ret == 0)
1357                         ret = -ETIMEDOUT;
1358                 goto out;
1359         }
1360
1361         /*
1362          * By now we've committed to the send.  We reuse rds_send_worker()
1363          * to retry sends in the rds thread if the transport asks us to.
1364          */
1365         rds_stats_inc(s_send_queued);
1366
1367         ret = rds_send_xmit(cpath);
1368         if (ret == -ENOMEM || ret == -EAGAIN) {
1369                 ret = 0;
1370                 rcu_read_lock();
1371                 if (rds_destroy_pending(cpath->cp_conn))
1372                         ret = -ENETUNREACH;
1373                 else
1374                         queue_delayed_work(rds_wq, &cpath->cp_send_w, 1);
1375                 rcu_read_unlock();
1376         }
1377         if (ret)
1378                 goto out;
1379         rds_message_put(rm);
1380
1381         for (ind = 0; ind < vct.indx; ind++)
1382                 kfree(vct.vec[ind].iov);
1383         kfree(vct.vec);
1384
1385         return payload_len;
1386
1387 out:
1388         for (ind = 0; ind < vct.indx; ind++)
1389                 kfree(vct.vec[ind].iov);
1390         kfree(vct.vec);
1391
1392         /* If the user included a RDMA_MAP cmsg, we allocated a MR on the fly.
1393          * If the sendmsg goes through, we keep the MR. If it fails with EAGAIN
1394          * or in any other way, we need to destroy the MR again */
1395         if (allocated_mr)
1396                 rds_rdma_unuse(rs, rds_rdma_cookie_key(rm->m_rdma_cookie), 1);
1397
1398         if (rm)
1399                 rds_message_put(rm);
1400         return ret;
1401 }
1402
1403 /*
1404  * send out a probe. Can be shared by rds_send_ping,
1405  * rds_send_pong, rds_send_hb.
1406  * rds_send_hb should use h_flags
1407  *   RDS_FLAG_HB_PING|RDS_FLAG_ACK_REQUIRED
1408  * or
1409  *   RDS_FLAG_HB_PONG|RDS_FLAG_ACK_REQUIRED
1410  */
1411 static int
1412 rds_send_probe(struct rds_conn_path *cp, __be16 sport,
1413                __be16 dport, u8 h_flags)
1414 {
1415         struct rds_message *rm;
1416         unsigned long flags;
1417         int ret = 0;
1418
1419         rm = rds_message_alloc(0, GFP_ATOMIC);
1420         if (!rm) {
1421                 ret = -ENOMEM;
1422                 goto out;
1423         }
1424
1425         rm->m_daddr = cp->cp_conn->c_faddr;
1426         rm->data.op_active = 1;
1427
1428         rds_conn_path_connect_if_down(cp);
1429
1430         ret = rds_cong_wait(cp->cp_conn->c_fcong, dport, 1, NULL);
1431         if (ret)
1432                 goto out;
1433
1434         spin_lock_irqsave(&cp->cp_lock, flags);
1435         list_add_tail(&rm->m_conn_item, &cp->cp_send_queue);
1436         set_bit(RDS_MSG_ON_CONN, &rm->m_flags);
1437         rds_message_addref(rm);
1438         rm->m_inc.i_conn = cp->cp_conn;
1439         rm->m_inc.i_conn_path = cp;
1440
1441         rds_message_populate_header(&rm->m_inc.i_hdr, sport, dport,
1442                                     cp->cp_next_tx_seq);
1443         rm->m_inc.i_hdr.h_flags |= h_flags;
1444         cp->cp_next_tx_seq++;
1445
1446         if (RDS_HS_PROBE(be16_to_cpu(sport), be16_to_cpu(dport)) &&
1447             cp->cp_conn->c_trans->t_mp_capable) {
1448                 u16 npaths = cpu_to_be16(RDS_MPATH_WORKERS);
1449                 u32 my_gen_num = cpu_to_be32(cp->cp_conn->c_my_gen_num);
1450
1451                 rds_message_add_extension(&rm->m_inc.i_hdr,
1452                                           RDS_EXTHDR_NPATHS, &npaths,
1453                                           sizeof(npaths));
1454                 rds_message_add_extension(&rm->m_inc.i_hdr,
1455                                           RDS_EXTHDR_GEN_NUM,
1456                                           &my_gen_num,
1457                                           sizeof(u32));
1458         }
1459         spin_unlock_irqrestore(&cp->cp_lock, flags);
1460
1461         rds_stats_inc(s_send_queued);
1462         rds_stats_inc(s_send_pong);
1463
1464         /* schedule the send work on rds_wq */
1465         rcu_read_lock();
1466         if (!rds_destroy_pending(cp->cp_conn))
1467                 queue_delayed_work(rds_wq, &cp->cp_send_w, 1);
1468         rcu_read_unlock();
1469
1470         rds_message_put(rm);
1471         return 0;
1472
1473 out:
1474         if (rm)
1475                 rds_message_put(rm);
1476         return ret;
1477 }
1478
1479 int
1480 rds_send_pong(struct rds_conn_path *cp, __be16 dport)
1481 {
1482         return rds_send_probe(cp, 0, dport, 0);
1483 }
1484
1485 void
1486 rds_send_ping(struct rds_connection *conn, int cp_index)
1487 {
1488         unsigned long flags;
1489         struct rds_conn_path *cp = &conn->c_path[cp_index];
1490
1491         spin_lock_irqsave(&cp->cp_lock, flags);
1492         if (conn->c_ping_triggered) {
1493                 spin_unlock_irqrestore(&cp->cp_lock, flags);
1494                 return;
1495         }
1496         conn->c_ping_triggered = 1;
1497         spin_unlock_irqrestore(&cp->cp_lock, flags);
1498         rds_send_probe(cp, cpu_to_be16(RDS_FLAG_PROBE_PORT), 0, 0);
1499 }
1500 EXPORT_SYMBOL_GPL(rds_send_ping);